CGI or AI product images: which do shoppers trust?
Shoppers do not distrust computer-made images as such. They distrust being misled. Surveys show that most people want to know when AI made an image, and in at least one experiment, ads labeled as AI-made scored lower on trust and purchase intent, most of all for expensive products. CGI built from the real product’s files sits on the other side of that line, but it only keeps that trust if the product in the image looks exactly like the one in the box.
What do surveys say about AI images and trust?
People want disclosure, and their tolerance for heavy AI use appears to be falling.
Three recent sources, each measuring something slightly different:
| Source | Who was asked | Key finding |
|---|---|---|
| Getty Images, Building Trust in the Age of AI (released 2024, data collected 2022 to 2023) | Surveys of 7,500 adults each across 25 countries | Almost 90% want to know if an image was created using AI; 98% say authentic images and video are pivotal to trust; 76% say it is getting hard to tell whether an image is real |
| Fractl (Q2 2026) | 1,008 US consumers | 40% say heavy AI use would lower their trust in a favorite brand, up from 20% in 2025; 90% want AI images labeled |
| Israfilzade, Equilibrium (December 2025) | Experiment with low- and high-involvement products (snacks and laptops) | Ads labeled as human-made got higher trust and purchase intent even when they were AI-made; AI labels hurt most for high-involvement products |
Read these with care. Getty sells photography and Fractl is a marketing agency, so both have a point of view. Surveys measure what people say, not what they buy. Still, the direction is consistent: people react to whether they feel informed, and a visible AI label costs some trust.
Why does this matter more for luxury and beauty?
Because the price is high and the product itself is the promise.
The experiment above found the strongest effect for high-involvement products such as laptops, the kind of purchase people think about before they buy. A fragrance, a serum or a watch fits that description better than a snack. Getty’s report also noted that people feel less favorably toward brands that use AI-generated visuals to create people or products, which is exactly what a product image does.
Beauty adds a second layer. Texture, color and visible effect are claims, and advertising regulators watch them. The UK’s ASA warns against production techniques that exaggerate what a cosmetic can do, whatever the tool. Our guide on CGI or photography for a beauty campaign covers where that line usually sits.
Is CGI “AI” in the eyes of the law and of shoppers?
Under the EU AI Act, a rendered image is not treated as AI output. For shoppers, it depends on whether the product looks real and correct.
The European Commission’s guidelines on the AI Act name a rendered frame as an example of content outside the AI marking rule, because it comes from calculation rather than AI generation. Our article on the EU AI Act and CGI explains the details. In a CGI production, artists build the bottle from its CAD files, apply the real label artwork and compare the result with a physical sample. The set and background may be invented, but the product itself traces back to its real files.
A generated image works differently. The model predicts plausible pixels based on what it has seen, and details such as label text, cap shape or liquid color can drift from the real thing. Since most people cannot tell how an image was made, they judge it on something else: whether the product that arrives matches the picture.
Where does generative AI go wrong with products?
In the details that carry the brand: text, proportions, materials and consistency.
The usual weak points are:
- Label typography and small text, including legal lines and ingredient names
- Proportions, such as cap height, shoulder shape or the cut of a stone
- Materials, for example how thick the glass looks or how a metal reflects
- Liquid color and fill level, which shift easily between images
- Consistency across shots, formats and variants of the same range
- Motion, where an object has to keep its shape and intent over several seconds
Not Important Studio showed the last point in a short experiment. It asked an AI tool to extend a single rendered frame into video. The motion held for a couple of seconds and then fell apart, which the studio described as a lack of intent and a reminder that some movements are easier to do by hand than to keep prompting for.
Does that mean AI has no place in product imagery?
No. What matters is which part of the image AI touches.
| Use of AI | Trust risk | EU labeling view, based on the Commission’s guidelines |
|---|---|---|
| Noise reduction or rescaling of a CGI render | Low | Treated as standard editing, outside the marking rule |
| Generated background behind a real or CGI product | Low to moderate | Not a deep fake, provided the ad does not mislead about the product |
| Product generated or altered by AI | High | Can be a deep fake if it could mislead about how the product looks or works |
| Generated people demonstrating the product | High | Likely to need disclosure |
This is our reading, not legal advice. The safe pattern is the same in both columns: keep AI away from the product, its texture and its effect, and be ready to say where it was used.
What should you ask before choosing?
Six questions settle most of it:
- Is the product built from our own files, or generated?
- Which elements, if any, were made with AI, and with which tools?
- Will the product look identical across every shot, format and variant?
- How and when do we compare the result with a real sample?
- If AI was used, what is our disclosure plan?
- Who owns the final images and the product model?
Put these in the brief from the start. Our brief checklist shows where they fit alongside product files and approvals.
Frequently asked questions
Can shoppers tell CGI from photography?
Often not. In Getty Images' research, 76% of people said it is getting hard to tell whether an image is real. What shoppers do notice is when the product they receive does not match the picture.
Do we have to disclose that an image is CGI?
Under the EU AI Act, a rendered image is not treated as AI output, so the AI label does not apply to it as such. Advertising rules against misleading claims still apply, whatever the tool.
Is a CGI product on an AI-generated background acceptable?
It can be. The European Commission's guidelines treat a real product in an AI-generated setting as not a deep fake, provided the ad does not mislead about the product. Check the specific case with your studio and legal team.
Do these surveys apply to our market?
Treat them as a direction, not a forecast. Fractl surveyed US consumers only, while Getty's research covered 25 countries, and both measure stated attitudes rather than actual buying.
Sources
- Getty Images press release: Nearly 90% of consumers want transparency on AI images (30 April 2024)
- Fractl: AI statistics, consumer and marketer survey (Q2 2026)
- Israfilzade, K.: AI-generated versus human-created advertising, effects on consumer trust and purchase intent. Equilibrium 20(4), December 2025
- European Commission: Guidelines on transparency obligations for providers and deployers of AI systems (July 2026)
- ASA: Cosmetics, the use of production techniques